Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

I gave Qwen 3.8 27B a reverse-engineering job and it finished in 30 minutes

xda-developers.com

Post reply on HN
Print view View original

141–150 of 159 posts

Re: I gave Qwen 3.8 27B a reverse-engineering job and it finished in 30 minutes

#141

Post by jnwatson »

Refefer wrote: post #110
jnwatson wrote: post #100

I can't get Qwen 3.8 27B to do a simple code review on a fairly basic Python file. With thinking on it just ruminates forever and with thinking off it gives obviously bad borderline hallucinating advice. Edit: I tried again with the 2.4T model and it still ruminates to death, but with thinking turned off, it generated genuinely useful advice. Edit2: adding --reasoning-budget 8000 --reasoning-budget-message "Reasoning…

One of the big learnings from 3.8 27b is adding reasoning budget really hurts the model. you need to let it spin for as many thinking tokens as it wants to to get it out. Another big takeaway is reasoning effort set to low doesn't save you tokens: low is pretty uncertain about things so it ends up thinking more (you can find some tests from folks on youtube). The final question, as always, is what quant are you runni…

It literally ran forever without a reasoning budget. I tried even the 2T model and cut it off after a half hour. This is to review a few hundred-line source file. It was consistent behavior from 2T to vanilla 27B to my ablated distilled version.

Re: I gave Qwen 3.8 27B a reverse-engineering job and it finished in 30 minutes

#142

Post by jnwatson »

nialv7 wrote: post #136
jnwatson wrote: post #100

I can't get Qwen 3.8 27B to do a simple code review on a fairly basic Python file. With thinking on it just ruminates forever and with thinking off it gives obviously bad borderline hallucinating advice. Edit: I tried again with the 2.4T model and it still ruminates to death, but with thinking turned off, it generated genuinely useful advice. Edit2: adding --reasoning-budget 8000 --reasoning-budget-message "Reasoning…

how are you running the 2.4t model locally if you don't mind me asking

Oh I didn't; only a distilled 27B model ran locally. This was just to figure out if distilling caused the problem.

Re: I gave Qwen 3.8 27B a reverse-engineering job and it finished in 30 minutes

#143

Post by Zambyte »

naasking wrote: post #130
Zambyte wrote: post #80

Earlier quoted context omitted.

Blackmail, incitement, libel, and basically anything that comes up if you search "computer crimes".

Generating blackmail is not illegal, using it to blackmail someone is. Generating libel is not illegal, publishing it publicly is not illegal either although you can be sued over it. Generating worms and computer viruses is not illegal last I checked, but disseminating them is.

LLMs have been able to make the leap between generating and sending for years now. I'm not sure what your point is.

Re: I gave Qwen 3.8 27B a reverse-engineering job and it finished in 30 minutes

#144

Post by naasking »

Zambyte wrote: post #143
naasking wrote: post #130

Earlier quoted context omitted.

Generating blackmail is not illegal, using it to blackmail someone is. Generating libel is not illegal, publishing it publicly is not illegal either although you can be sued over it. Generating worms and computer viruses is not illegal last I checked, but disseminating them is.

LLMs have been able to make the leap between generating and sending for years now. I'm not sure what your point is.

My point is that generating is not illegal but sending is. Your whole point hinges on legality, so the distinction between legal and illegal seems pretty key.

Re: I gave Qwen 3.8 27B a reverse-engineering job and it finished in 30 minutes

#145

Post by InvertedRhodium »

naasking wrote: post #129
jszymborski wrote: post #85

Earlier quoted context omitted.

The weights aren't deleted, it's just additional fine tuning, is my understanding.

There is no question model quality is degraded by this though.

It's altered, sure. I think inherent degradation is a step too far though.

Re: I gave Qwen 3.8 27B a reverse-engineering job and it finished in 30 minutes

#146

Post by mycall »

andai wrote: post #56
VulgarExigency wrote: post #5

> The first attempt at recovering the key was wrong in a very specific way; it produced a working key and the signature check passed, but a hash the binary computes as an integrity check didn't match. In my experience, most models would have called it done and left it at that, but Qwen 3.8 27B didn't do that. Instead, it highlighted the mismatch, went back to the drawing board, and kept going until the value matched…

Yeah, about a year ago the labs figured out that effective intelligence is a function of persistence as much as anything else. So the models started getting scary persistent late last year, and the trend has continued. There was another jump a few months ago.

Is persistence the opposite side of attention (all you need)?

Re: I gave Qwen 3.8 27B a reverse-engineering job and it finished in 30 minutes

#147

Post by le-mark »

naasking wrote: post #126
egl2020 wrote: post #113

Earlier quoted context omitted.

Any idea how being persistent is trained? I've noticed that telling an LLM that it needs to think some more sometimes produces better results, but the claim here is that "they are very persistent" and "...kept going...".

It's from work like this: https://arxiv.org/abs/2309.11495 A RL pipeline can reinforce verification behaviour even better than simple prompting.

They can also be more effective reviewing than generating.

Re: I gave Qwen 3.8 27B a reverse-engineering job and it finished in 30 minutes

#148

Post by Neywiny »

verdverm wrote: post #118
colingauvin wrote: post #82

Earlier quoted context omitted.

You can also just do read only access through most mcp servers for functionally no difference

if you download them, you can remove all internet access for the model going forward on the task, don't want to http proxy like OpenAi now do we?

Yeah that's more my point. I've seen the logs for what these things do when I leave them unattended. I wouldn't trust anything to go out to the open network, especially with my credentials. Maybe mcp server to an offline database.

Re: I gave Qwen 3.8 27B a reverse-engineering job and it finished in 30 minutes

#149

Post by andai »

mycall wrote: post #146
andai wrote: post #56

Earlier quoted context omitted.

Yeah, about a year ago the labs figured out that effective intelligence is a function of persistence as much as anything else. So the models started getting scary persistent late last year, and the trend has continued. There was another jump a few months ago.

Is persistence the opposite side of attention (all you need)?

AGI ≈ Artificial Stupidity × Infinite Persistence

Ralph is all you need!

https://ghuntley.com/ralph/

Re: I gave Qwen 3.8 27B a reverse-engineering job and it finished in 30 minutes

#150

Post by KennyBlanken »

VulgarExigency wrote: post #5

> The first attempt at recovering the key was wrong in a very specific way; it produced a working key and the signature check passed, but a hash the binary computes as an integrity check didn't match. In my experience, most models would have called it done and left it at that, but Qwen 3.8 27B didn't do that. Instead, it highlighted the mismatch, went back to the drawing board, and kept going until the value matched…

Unfortunately this "persistence" seems to trigger a lot of issues with agents getting stuck in loops.
Post reply on HN